Agency Capacity Planning When AI Does the Work

The old capacity question was "how many people do we need to produce the work?" The new one is "how much can we review and stand behind?" When AI does the production, your constraint stops being hands on keyboards and becomes judgment on outputs. Agency capacity planning in the AI era means counting review hours and quality gates, not headcount. Get this wrong and you either turn down work you could easily deliver or you drown in output you cannot vouch for.

Why does AI break traditional capacity planning?

Because traditional planning assumes production is the expensive, scarce resource. You staffed to the amount of work you could physically make. Ten designers made roughly ten designers' worth of output, and to double that you hired ten more.

AI collapses the cost of production. The same person can now generate the raw output of a whole team. So if you keep planning capacity around production, your numbers are nonsense, they describe a bottleneck that no longer exists. The real bottleneck moved. It is now the human review, the strategic direction, and the accountability for what ships. Those did not get cheaper. If anything they got more valuable, because someone still has to stand behind every deliverable that goes out with your name on it.

I have written about this shift from the other direction in how AI agents let one person run twenty companies. The pattern is identical: production stops being the limit, and orchestration becomes the limit.

What should you actually measure now?

Measure the scarce thing, which is qualified human attention. Three numbers.

Review capacity: how many deliverables can a person actually check, correct, and approve to your standard per week? This is now your true throughput ceiling. Judgment load: how many accounts can one strategist genuinely think about at once before the thinking gets shallow? Accountability surface: how much output can you put your name on and defend if a client challenges it?

Notice that none of these is "how much can we produce." Production is assumed to be abundant. The plan is built around the constraint that remains, which is trust in the output. An AI agency operating system that handles the production frees your people to spend their entire capacity on review and strategy, which is exactly where you want it spent.

How do you plan capacity around review, not production?

Start by separating the two kinds of work. Production work is what AI can do at near-zero marginal cost. Review work is what a human must do to make it safe to ship. Plan your capacity around the second column only.

If a strategist can properly oversee eight accounts, then eight is your capacity per strategist, no matter how much raw production the system can generate. Adding more production capacity does nothing if the review capacity is full. This is counterintuitive for agencies used to thinking that more output equals more capacity. More output you cannot review is not capacity, it is liability.

This is why I argue an agency should ship a system, not a campaign. A system is designed so the review is efficient and the quality gates are built in, which raises how many accounts one person can safely stand behind. Improve the review process and you improve your real capacity without hiring anyone.

Does this mean you never hire again?

Not never, but you hire differently. You do not hire to produce more, because production is no longer scarce. You hire to review more, direct more, and be accountable for more. The role shifts from maker to overseer, from doing the work to judging the work.

That is a genuinely different hire. It rewards taste, judgment, and the ability to catch what is wrong, not raw production speed. Some of your best producers will make great reviewers. Some will not, because catching errors and being decisive about quality is a distinct skill. Plan for that. I make the case for staying lean as long as possible in how to run a lean agency with AI, and the short version is that most agencies add reviewers far too early, before they have squeezed the leverage out of the system they already have.

The bottom line

Capacity planning used to be a headcount math problem. Now it is a judgment and accountability problem. Count what is scarce, which is qualified review and strategic attention, and plan around that. Stop counting production, because production is no longer the thing that runs out. The agencies that keep planning around bodies will over-hire and watch their margins sink. The ones that plan around review capacity will take on more clients with fewer people and stand behind every deliverable they ship. The work got cheap. The judgment got expensive. Build your capacity plan around the expensive thing.